[Lustre-discuss] OST crash recovery problem

[Heiko Schroeter]

Am Dienstag, 26. August 2008 15:35:38 schrieb Jeremy Mann:
> Mag Gam wrote:
> > LOL... I am in the same situation, I want to see what problems other
> > people have so I can try to help them and I can further avoid it. I am
> > a big proponent of "your problems are my problems" :-)
>
> When we first implemented Lustre, I had several learning curves with
> OSTs going down, drives failing... Eventually we settled on backing up
> the Lustre filesystem to a backup array in the case a OST would fail. It
> does take some work, but we find that rebuilding Lustre after an OST
> failure works for us.

Hm, we have 52TB OST space so far and more than 25TB are coming within the 
next days. So no backup space there ...

I can live with an OST breaking down. Since we do stripe a single file onto a 
single OST and the lustre works as a 'fast' data archive there is no problem 
to recreate the data on a single OST.

But in this case the faulty OST couldn't be removed from the lustre system at 
all ! (Or at least we couldn't do it ...)

That is the deeper reason why we did a new setup.
Sorry if i haven't said this earlier.

Heiko